Elon Musk has capped Tesla employees' spending on AI at $200 (£150) a week as companies seek to rein in runaway bills.

Staff at the electric vehicle manufacturer have been told that their use of the technology will be limited from Monday.

The cap does not apply to Grok, Mr Musk's own AI system, which has struggled compared to rivals such as Anthropic's Claude, ChatGPT and Google's Gemini.

Many large companies have pushed their employees to use AI as much as possible hoping it will make them more productive.

Mr Musk said earlier this year that AI would allow "output" per Tesla worker to get "nutty high".

But the push to use AI has led to soaring bills, as the cost of using these systems rises with demand. There are fears that staff are also wastefully using AI to carry out menial tasks in an attempt to demonstrate they are using it.

Several companies are now seeking to rein in staff use. Uber, which had told employees to use AI as much as possible, recently limited usage to $1,500 per month, while Meta, Walmart and Coinbase have all said they will introduce caps.

Capping spending could force staff to use AI only for important tasks or encourage them to use cheaper AI systems.

Tesla's $200-a-week cap was first reported by the tech news website The Information. Staff will need to receive special permission to go above the limit.

The limits mark an about-turn from the "tokenmaxxing" trends in which staff were measured by how many tokens, a unit of AI usage, they were consuming.

A pull-back on corporate AI usage has stoked fears of a potential stock market crash. Heavy spending on AI infrastructure by labs has helped propel markets to new heights, but this investment is predicated on rapid adoption of the technology.

Responding to a message on X earlier this week predicting an imminent tech crash, Mr Musk wrote: "There are always momentary dips, even in a rapidly growing economy. The productivity gains from AI and robotics are so enormous, however, that the macro trend is overwhelmingly up."

Mr Musk's own AI company xAI, which is owned by SpaceX, has struggled to match rival chatbots, especially when it comes to the coding tasks that Claude and ChatGPT are widely used for. The billionaire has said the system needs to be "rebuilt from the ground up".

SpaceX recently agreed to acquire AI coding start-up Cursor for $60bn. The company has its own AI system, Composer, which Mr Musk has encouraged Tesla staff to use.

Mr Musk is determined to make Tesla an AI leader through its use of driverless car technology and the company's Optimus robots. The push to take advantage of the technology comes amid growing competition for its EVs from cheap Chinese electric cars.

In just a short amount of time, corporate policy on "tokenmaxxing" — or excessively using AI models to inflate usage on internal leaderboards — has already peaked and valleyed. After seeing their bill for AI tokens, many companies have shifted toward a more efficient outlook on AI spending.

Tabs CEO Ali Hussain discusses what's changed within corporate AI pricing and what this pivot may mean for OpenAI (OPAI.PVT) and Anthropic (ANTH.PVT) ahead of their mega-IPOs.

Palantir Technologies CEO Alex Karp said Wednesday that OpenAI and Anthropic have fundamentally mispriced their AI services, arguing the token-based model has left enterprise customers frustrated and empty-handed.

On CNBC's "Squawk Box," Karp put it bluntly: "Something has gone completely wrong." He said the prevailing attitude among U.S. businesses has become one of resignation — that they will burn through tokens, generate no value, and hand over their intellectual property in the process.

He stopped short of personally targeting AI executives — quipping that he finds private debates with Anthropic CEO Dario Amodei entertaining — while nonetheless insisting that the underlying models have been "completely, irresponsibly, oversold." In private conversations, he said, executives have confided that token spending yields nothing useful, and that engaging with AI labs puts their proprietary data and market advantages at risk.

"Data retention is your treasure. Transfer it at your own peril," Palantir's recently published nine-point manifesto on AI sovereignty reads.

After co-anchor Becky Quick suggested he seemed furious, Karp pushed back with a reframe: The frustration he was expressing, he said, belonged not to him personally but to corporate America broadly — "the voice of American business that is being channeled through me." To illustrate how widespread the discontent runs, Karp suggested anyone skeptical should ring a CEO and relay what they had just heard on air — the response, he predicted, would be that the executive was at least twice as livid as Karp himself appeared.

Karp pointed to open weight models as an alternative path for enterprises seeking more control. Speaking to CNBC, Karp framed the Nvidia alignment as a shared belief in customer ownership: technically sophisticated clients, he said, want sovereignty over their compute, models, data, and competitive edge — not a setup where those assets quietly migrate elsewhere. The remarks followed a deal unveiled earlier this week in which Palantir and Nvidia agreed to bring Nvidia's open Nemotron models into U.S. government agencies and critical infrastructure, pairing Nvidia's compute capabilities with Palantir's data and AI stack across classified and air-gapped settings.

Karp's remarks land amid a broader pullback from the tokenmaxxing era, when some of the largest companies in the world prioritized burning as many AI tokens as possible with little attention to returns. Companies including Uber and Microsoft have capped or restricted employee access to expensive AI coding tools after budgets blew out. A growing number of enterprise customers have also been cutting spending on OpenAI and Anthropic, switching to cheaper alternatives and demanding clearer returns on investment.

By Aditya Soni

June 29 (Reuters) - Silicon Valley's powerful and pricey AI models have been a necessity for businesses looking to future-proof themselves. But now a growing number of tech CEOs are arguing that cheaper options would be crucial for their wider adoption.

Top executives such as Microsoft's Satya Nadella, Palo Alto Networks' Nikesh Arora ‌and Coinbase Global's Brian Armstrong have said smaller, cheaper models can handle a big share of corporate needs.

This view is the result of a reassessment within companies ‌that until recently encouraged heavy use of AI tools, often treating rising consumption as a proxy for productivity, dubbed "tokenmaxxing". Now, those bills are starting to bite.

Prices of tokens - the units used to measure AI usage - are falling, but the ​cost of completing a task is rising as AI firms shift from flat subscriptions to usage-based pricing. That is leaving companies with unpredictable and often higher bills as usage per task becomes harder to estimate.

Uber, for instance, burned through its entire 2026 AI budget in just four months after employees rushed to adopt AI coding tools, forcing management to cap usage, according to reports.

"Changing the license model caught a lot of people by surprise," said Harold Byun, CEO of BlueRock, a startup that helps companies run AI systems safely. "Immediately after that, we had a number of reports from customers that ‌we're seeing a 20% to 30% spike in terms of over-budgeting."

BUSINESSES ⁠FRET OVER HUGE BILLS

As companies use AI more, their costs are surging beyond initial estimates as tasks now involve more steps, more data and longer inputs.

Gartner estimates AI coding costs will surpass the average developer's salary by 2028, while a survey by the research firm found three-quarters of executives see ⁠tech budgets rising this year, with nearly half of them projecting double-digit jumps.

That has led businesses to embrace cheaper models and turn to routing tools such as OpenRouter, an AI marketplace, as they seek to assign tasks to the most cost-effective system while reserving premium models for complex work such as coding.

Open-source tokens processed on OpenRouter jumped to 65% in June from 34% in January, according to a Citi note.

That ​should ​benefit open-source model makers such as China's DeepSeek, which have won wide adoption among startups but struggled to ​break into large businesses due to security concerns.

"If you want to win ‌enterprise, you should be forward pricing tokens," Palo Alto Network's Arora wrote on X last week, urging AI labs to charge customers today at the lower rates that tokens are expected to command in a few years.

The era of tokenmaxxing is over. After the AI industry encouraged companies to max out their AI budgets earlier this year, and some companies even built employee leaderboards to encourage internal AI usage — they are now realizing just how easy it is to spend huge sums of money on AI and get little in return.

We now appear to be entering the era of token rationing.

Recent news has been rife with stories about AI cutbacks and now 404 Media reports that consulting firm Accenture has been attempting to stop its employees from depleting its token reserves by using AI to do basic tasks — like converting PDFs into presentation slides.

The cutbacks take place not long after Accenture threatened that employees would "risk losing out on promotions" if they didn't use AI, 404 writes.

404's reporting is based on leaked audio from a recent internal meeting involving Accenture's agentic AI strategy lead, Justice Kwak.

"We're hitting this inflection point where AI is becoming material to the cost structure," Kwak says. "Spend is becoming very unpredictable; and leadership, especially at the CFO, COO, and CIO level, are still asking the question of whether they're getting value from what we're spending on in the context of AI."

The cost of tokens has thrown into doubt the AI business model — as evidenced by what's being called the "AI selloff" which has battered some AI-dependent businesses the last few days, especially memory chip makers. The AI industry has reached the stage where it can't just be exciting and new anymore. It has to prove its worth.

This story was originally published on CIO Dive. To receive daily news and insights, subscribe to our free daily CIO Dive newsletter.

Dive Brief:

• The rise in price of tokens will lead AI coding to become more expensive than an average developer's salary by 2028, according to a Gartner report released Wednesday. The shift will happen as consumption-based pricing overtakes subscription models among key providers.

• Changing pricing models is also making AI costs a highly variable figure, keeping enterprise tech leaders from accurately forecasting and controlling spending, the report said. Vendors frequently lack transparency into how token consumption is calculated and billed. 

• The cost structure is changing as enterprises still struggle to find maturity in AI projects and measure their business impact, said Nitish Tyagi, senior principal analyst at Gartner, in a statement. "Software engineering leaders are increasingly concerned as token-driven AI spend becomes harder to justify, with budgets often being depleted earlier than expected."

Dive Insight:

Adopting AI into software workflows has become the default, leading employees to spend less time writing code and more time managing AI outputs. 

As AI use proliferates the enterprise, the cost is adding up, especially in engineering departments, Gartner found. Token overspending was linked to how software engineering leaders govern usage, with many using ungoverned, autonomous agents in their workflows. 

"AI coding costs will continue to rise as infrastructure investment and profitability challenges push model pricing higher," Tyagi said. "At the same time, as more developers adopt AI tools, light users are expected to rapidly become mainstream users as familiarity and reliance increase, driving further growth in token consumption and overall spend." 

Few enterprises have a clear strategy with defined goals and outcomes for their AI projects, according to Altimetrik data published in April, but they choose to forge ahead rather than miss out on the AI moment. Just over one-quarter of enterprise C-suite leaders reported having complete, real-time visibility into what their AI systems cost to operate, a KPMG report released Wednesday found. 

It's likely many agents are working on tasks in the background for days, without leaders' knowledge or audits, Rahsaan Shears, AI enterprise transformation leader at KPMG LLP, told CIO Dive in an email. 

"The CFO does not see it. The CIO may not either," Shears said. "That is enterprise AI economics right now: costs compounding inside workflows no one has fully instrumented."